The invention relates to clocking circuits, and more particularly to circuits for generating a clock signal and a symmetrical complement of the clock signal.
Clock signals are commonly used in digital circuits, including circuits used in memory devices, to control the timing at which various events occur. In some cases, a single clock signal is used. However, in other cases, it is necessary to use both the clock signal and the complement of the clock signal. The signals are typically generated by applying a clock signal to a phase splitter, which then generates a clock signal and its complement for use by the digital circuit.
It is important that the clock signal and its complement be symmetrical, i.e., the edges of both signals be substantially aligned and have the same slew rate. The clock signal and its complement generated by an ideal phase splitter would have a 50 percent duty cycle, equal rise and fall times and they would be exactly 180 degrees out of phase from each other. In practice, the ideal is rarely achieved for a variety of reasons, including process variations in the fabrication of the phase splitter. For example, for a CMOS process, one pass of the process may result in an inverter having a fast NMOS transistor and a slow PMOS transistor, and another pass of process may result in the inverter having a slow NMOS transistor and a fast PMOS transistor. As a result, the inverters will respond differently to incoming clock signal, and the respective clock signals generated by the inverters will not be symmetrical.
A conventional phase splitter 10 is illustrated in FIG. 1. The phase splitter 10 includes two branches 12, 14, one of which generates a signal OUT and the other which generates its complement OUT*. The second branch 14 consists of three inverters 20, 22, 24. Since there are an odd number of inverters in the second branch 14, the complementary output signal OUT* is the complement of the input signal CLK, but delayed in time by the sum of the propagation delays through each of the inverters 20-24.
The first branch 12 consists of two inverters 30, 32 and a capacitor 34 connected to the output of the first inverter 30. The size of the capacitor 34 is selected to delay the coupling of all of signals from the output of the first inverter 30 to the input of the second inverter 32 by an amount corresponding to the difference between the delay of the three inverters 20-24 and the two inverters 30, 32. As a result, the OUT signal and the OUT* signal are theoretically 180 degrees out of phase with each other.
In practice, however, the OUT and OUT* may not be entirely symmetrical for several reasons. For example, although the capacitor 34 compensates for the delay of the extra inverter in the second branch 14, it also reduces the slew rate of the signal applied to the input of the inverter 32. As a result, the slew rate of the signal applied to the inverter 32 is substantially slower than the slew rate of the signal applied to the inverter 24. This difference in slew rates causes the rise and fall times of the signals OUT and OUT* to differ substantially from each other.
Proposals have been made to modify the prior art phase splitter 10 shown in FIG. 1 by dispensing with the capacitor 34 and instead adjusting the delay of each of the inverters 20-24, 30, 32 to achieve substantially the same result. More specifically, the inverters 20, 24 and 30 may be designed so that the sum of the delays through the inverters 20, 24 is equal to the delay through the inverter 30. The inverters 22 and 32 are then designed so that they have equal propagation delays. As a result, the signals OUT and OUT* are, in theory, symmetrical. Again, in practice, the signals are anything but symmetrical for several reasons. For example, the inverters 20, 24 must be relatively fast so that the sum of their delays is equal to the delay of the inverter 30. The high-speed of the inverter 24 causes it to have a relatively high slew rate. In order for the slew rate of the OUT signal to match the slew rate of the OUT* signal, the transistors used in the inverter 32 must be relatively large. However, the inverter 30 must be fairly slow to achieve the required delay, and, as a result, its output signal has a relatively low slew rate. The low slew rate of the inverter 30 makes it all the more difficult for the output of the inverter 32 to match the output of the inverter 24 so that OUT and OUT* will have the same rise and fall times. If the slew rate of the inverter 24 is decreased to match the slew rate of the inverter 32, the speed of the inverter 32 will also be reduced. As a result, is necessary to increase the speed of the inverter 20 by a commensurate amount, thereby making the inverter 20 very large. In addition to consuming a relatively large area of the substrate, making the inverter 20 large decreases the input impedance of the inverter 20 making it difficult for other circuits (not shown) to drive the inverter 20.
A need therefore exists for a phase splitter that uses relatively little circuitry consuming relatively little area on a substrate that produces from a clock signal complementary signals that are substantially symmetrical in both phase and slew rate despite fabrication processing variations.
A phase splitter in accordance with the invention is operable to generate first and second complimentary output clock signals from an input clock signal. The phase splitter includes two branches receiving the input clock signal and generating the respective output clock signals. The first branch includes an even number of series connected inverters while the second branch includes an odd number of series connected inverters. In one aspect of the invention, an inverter is coupled between an output of an inverter in the first branch that is N number of inverters from the input clock signal and an output of an inverter in the second branch that is N+1 number of inverters from the input clock signal, where N is a positive integer, and where the added inverter""s input is on the second branch and its output is on the first branch. Cross-coupling signals between the first and second branches in this manner provides delay but subsequently has increased the slew rate of the signal applied to an input of one of the inverters. In another aspect of the invention, a first pair of diode-connected transistors are coupled between an output of one of the inverters in the first branch that is M inverters from the input clock signal, and M+P inverters in either branch, to make the phase splitter substantially insensitive to process variations. M is a positive integer and P is positive even number. For the same reason, a second pair of diode-connected transistors are coupled between an output of one of the inverters in the second branch that is M inverters from the input clock signal, and M+P inverters in either branch.